Point to point Control of a Brachiation Robot Based on Neural Network

Please enable javascript in your browser.

Babaei, Bashir | 2014

569 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 46953 (08)
University: Sharif University of Technology
Department: Mechanical Engineering
Advisor(s): Vossoughi, Gholamreza
Abstract:
• Brachiation robot is a kind of under-actuated robots. A brachiating robot is a type of a mobile arm that is capable of moving from branch to branch similar to a long-armed ape. The purpose of this thesis is control of a two link Brachiation robot. The purpose of this thesis is control of a two link Brachiation robot using neural network and geometry control. For this purpose a genetic algorithm based training Neural Network is used to produce a suitable path for the second link of the robot and then using input-output linearization method, the second link is controlled to follow the path. The simulations shows that the Nero Controller designed in this thesis is suitable. The controller can control the robot from each initial condition to a desired point. In fact the controller is independent from initial condition and it is robust from up to 10% inertial changes. The controller also can control the robot in presence of 1 KHz noise up to 10e-5 1/Hz power
Keywords:
Brachiation Robot ; Neural Network ; Geometric Control ; Genetic Algorithm ; Underactuated System ; Input-Output Linearization

No TOC